Codesota · Models · NV-Embed-v2NVIDIA2 results · 2 benchmarks
Model card

NV-Embed-v2.

NVIDIAopen-source7B paramsMistral-7B (LLM-based embedding)

Rank #1 on MTEB (56 tasks) as of Aug 2024 with 72.31 average score.

§ 01 · Benchmarks

Every benchmark NV-Embed-v2 has a recorded score for.

#BenchmarkArea · TaskMetricValueRankDateSource
01BEIRNatural Language Processing · Text Rankingndcg@1062.6%#1/42024-09-01source ↗
02MTEB LeaderboardNatural Language Processing · Feature Extractionavg-score72.3%#1/62024-09-01source ↗
Rank column shows this model’s position vs all other models scored on the same benchmark + metric (competitors after the slash). #1 in red means current SOTA. Sorted by rank, then newest result.
§ 02 · Strengths by area

Where NV-Embed-v2 actually performs.

Natural Language Processing
2
benchmarks
avg rank #1.0
§ 03 · Papers

1 paper with results for NV-Embed-v2.

  1. 2024-09-01· Natural Language Processing· 2 results

    NV-Embed: Improved Techniques for Training LLMs as Generalist Embedding Models

§ 04 · Related models

Other NVIDIA models scored on Codesota.

MambaVision-L2
241M params · 1 result
Nemotron 3 Nano (30B)
1 result
Nemotron 3 Super (120B)
1 result
Nemotron Nano V2 VL
1 result
Canary-1B
1B params · 0 results
FasterViT-6
0 results
RVT
0 results
RVT-2
0 results
§ 05 · Sources & freshness

Where these numbers come from.

arxiv
2
results
2 of 2 rows marked verified.